“Fill up your tax bracket” is the most common Roth conversion advice there is, and it skips the only part that’s hard: which bracket? The 12%? The 22%? Halfway into the 24% because it’s only two points more? I got tired of the hand-waving, so I ran my own Roth conversion optimizer 34 times, changing one thing at a time, and watched where it stopped.
Short version: it almost always stops exactly at a bracket edge, never somewhere in the middle. Which edge depends on a single comparison, and that comparison isn’t the one most people make.

The setup
I used the tool’s sample case: single, retiring at 62, $1.2 million in a traditional IRA, $300,000 in a brokerage account, $50,000 in a Roth, $60,000 a year of spending, $36,000 of Social Security starting at 70, a 5% real return, and a plan that runs to 90. Federal tax only, on 2026 brackets. No state tax, and IRMAA and the ACA subsidy switched off, because I wanted to see the brackets on their own.
Then I changed one input: the tax rate the leftover IRA will pay later, by you in your RMD years or by your heirs after you. I stepped it from 10% to 36%. Then I did the whole thing again with a $3 million IRA.
It stops at the edge, not in the middle
With the $1.2 million IRA and a future rate of 10% to 12%, the plan was the same every time: convert from 62 to 74 and stop with taxable income at exactly $50,400. That’s the top of the 12% bracket for a single filer in 2026. About $409,000 converted in total, and nothing more after that. From 13% up, the answer starts to move a step at a time: about $455,000 at 13%, $500,000 at 14%, $651,000 at 15% and $878,000 at 20%. Each extra point buys a few more years of converting past $50,400 and into the 22% bracket, mostly right at 62 and in the years after required distributions start at 75.
At 23% and above, the plan jumped again. It converted every year all the way to 90, using the 22% bracket up to $105,700 of taxable income, and converted $1,317,001 over the plan, which empties the IRA by 90. And from 23% all the way up to 36%, it made no further changes. It never touched the 24% bracket, because the room in the 22% bracket over those years was enough to finish the job.
The reason it lands on edges is simple once you see it. Inside a bracket, every dollar of conversion costs the same rate. If one dollar in the 12% bracket is worth converting, the whole rest of that bracket is too, so the optimizer takes all of it. It stops where the price changes, and the price changes at the edge. (Not every single year ends exactly on an edge. Spending, capital gains and Social Security move things around a bit. But the pattern is clear.)
The only question: is this rate lower than the later one?
That’s the whole rule. Fill a bracket if its rate is lower than the rate the same dollar would pay if you left it in the IRA. Stop at the edge of the first bracket where that isn’t true. The same rule holds for federal workers, since you can now do a Roth conversion inside the TSP.
The chart shows where it’s sharp and where it isn’t. Below 23%, the total for the $1.2 million IRA climbs gradually, because each extra point makes a few more years worth converting into the 22% bracket. Then going from 22% to 23% added more than $300,000 in one step, and from a 20% future rate to 23% the conversions grew about 1.5 times. Going from 23% to 36% changed nothing. At 22% it’s close to a wash: paying 22% now or 22% later leaves you with about the same money. So what matters most isn’t how precisely you guess your future rate. It’s which side of the next bracket rate that guess lands on.
“Later” includes your own RMDs
Here’s the part I didn’t expect. With the $3 million IRA and a future rate of just 10%, where you’d think converting at 22% is a terrible deal, the optimizer still converted about $1,097,000. It filled the 22% bracket every year from 62 to 74, and went into the 24% bracket for a few years after that.
The reason is the required minimum distributions. For someone born in 1964 they start at 75, and on a $3 million IRA they’re big: about $144,000 in the first year of that plan, growing from there. Those forced withdrawals push taxable income into the 24% bracket right away, into the 32% bracket by 84, and into the 35% bracket by 90. The 10% I entered was the rate on whatever is left at the end. But a lot of that IRA never makes it to the end. It comes out as RMDs at 24% to 35%. Converting at 22% in your sixties beats that easily.
So “compare today’s rate to your future rate” is right, but “future rate” means the highest rate that money would otherwise pay: your own rate in your RMD years, or your heirs’ rate, whichever it actually ends up hitting. With a small IRA, RMDs stay modest and the heirs’ rate decides. With a big one, your own RMD years decide, and you’ll want to fill more than feels comfortable.
At 25% and above, the $3 million plan converted the whole IRA, about $3.49 million, mostly through the 24% bracket. The 32% bracket never showed up, even at a 36% future rate: taxable income tops out at $201,775, the edge of the 24% bracket.
Your bracket may not be your rate
One more reason the edges in the runs above sit where they do. Once benefits start, each extra dollar of income can also make up to 85 cents of Social Security taxable, and a conversion in the 12% bracket can really cost 18% or even 22%. That’s what pulls the $1.2 million plan into the 22% bracket at a future rate as low as 13%: when I reran it with Social Security set to zero, it never went past $50,400 at a 13%, 15% or 20% future rate. The retirement withdrawal tax calculator shows this directly: it gives you your bracket, the real tax on your next $1,000, and how much room is left before the next edge.
So how full?
- Fill each bracket whose rate is below the highest rate that money would pay later, yours or your heirs’.
- Stop at the edge of the first bracket that isn’t. Not partway in.
- If your IRA is big enough that RMDs will push you into the 24% or 32% bracket anyway, “later” is your own seventies and eighties, not just your heirs.
- Check your real rate, not just your bracket, once Social Security starts.
For one year, that’s arithmetic you can do with the withdrawal tax calculator. Across a whole retirement it isn’t, because this year’s conversion changes next year’s balances, the RMDs at 75, and which account pays for everything. That’s what the optimizer is for. And if you’re near 63, turn on the Medicare surcharges under Advanced: the IRMAA cliffs are edges too, and they cost a lot more than a bracket.
These runs are one sample case with federal tax only, and the future rate is a guess you have to make yourself. This isn’t tax advice. Bring your numbers to a CPA before you convert anything real.
Updated November 2026: numbers rerun after I fixed a bug in the model (Social Security and other income now pay for spending, as they should).
Things that I use, like, and am affiliated with:
Mint Mobile offers great cell phone service for $15 flat, get $15 off using the link. Get discounted phones with service activation and no contract.
I never spend money before I check Mr Rebates or Rakuten to get cashbacks, rebates, discounts, coupons or cheaper gift cards.
